Underwater image restoration method based on adaptive brightness improvement and local image de-scattering

Abstract

The invention discloses an underwater image restoration method based on adaptive brightness improvement and local image de-scattering. The method comprises the following steps: acquiring an original underwater image; constructing an attenuation model and a brightness correction matrix of the original underwater image from the water surface to the scene; defining a brightness compensation function according to the attenuation model and the brightness correction matrix, decomposing the brightness compensation function into a plurality of sub-problems, solving the sub-problems by using an alternating direction multiplier method to obtain a value corresponding to brightness compensation, and obtaining an underwater image after brightness compensation; obtaining medium transmissivity and underwater environment light of R, G and B channels based on a DCP algorithm, and processing the underwater image after brightness compensation to obtain a restored image; according to the method, the problem of brightness degradation of the underwater image is concerned, so that the underwater image is closer to an image under real ambient light, and better balance is achieved in the aspects of processing fuzzy contrast and low brightness of different underwater images; and the underwater image can be well restored without any fine tuning of parameters.

Description

Technical Field

[0001] The invention relates to the technical field of digital image processing, and in particular to an underwater image restoration method based on adaptive brightness improvement and local image descattering. Background Art

[0002] The complex imaging environment of the ocean causes serious degradation of underwater images acquired by the optical vision system. The degradation of underwater images mainly includes: color deviation caused by light absorption, blurred details caused by forward scattering of light, and low contrast caused by backscattering of light, which causes the acquired underwater images to have problems such as color degradation, low contrast, and blurred details.

[0003] High-quality underwater images can provide useful support for extracting some valuable image content, which is of great significance to many marine applications such as underwater vehicles and underwater target detection. It is also beneficial to the development of application performance in remote sensing information extraction and target detection. In the past few decades, underwater image restoration and enhancement have attracted widespread attention from researchers. These works can be roughly divided into three categories: one is hardware-based methods, which usually use dedicated hardware such as polarization filters to obtain multiple images and use the complementary information between them to reconstruct clear underwater images; the second is non-physical model image enhancement methods, which adjust the pixel values ​​of the image to improve color, details and structure. For example, CN202010069967.0 discloses an image enhancement method based on color correction and deblurring. The proposed method effectively integrates the two stages of color correction and deblurring. CN201910083780.3 discloses an underwater image enhancement processing method and device. The third category is a restoration method based on a physical model, that is, the medium transmittance and underwater ambient light of the underwater scene are estimated based on a strong prior or assumption, and then the physical model is inverted to obtain a high-quality underwater image. These commonly used priors or assumptions include dark channel priors, underwater dark channel priors, generalized dark channel priors, etc. For example, CN202210354965.5 discloses a bright channel-based underwater image restoration method, and CN202110419038.2 provides an underwater image restoration method based on adaptive atmospheric light fusion.

[0004] Most existing model-based methods tend to estimate the global transmission through various priors and use the same global transmission method to reconstruct the R, G, and B channels of underwater images. However, data show that in underwater scenes, there are significant differences in the attenuation levels of the R, G, and B channels, resulting in unsatisfactory color restoration of these methods. Summary of the invention

[0005] The present invention provides an underwater image restoration method based on adaptive brightness improvement and local image descattering, so as to overcome the technical problem of unsatisfactory color restoration in the process of underwater image restoration using a priori method and global transmission method.

[0006] In order to achieve the above object, the technical solution of the present invention is:

[0007] An underwater image restoration method based on adaptive brightness improvement and local image descattering, comprising:

[0008] S1: Acquire an original underwater image, where the original underwater image is an RGB image;

[0009] S2: constructing an attenuation model and a brightness correction matrix of the original underwater image from the water surface to the scene;

[0010] S3: defining a brightness compensation function according to the attenuation model and the brightness correction matrix, decomposing the brightness compensation function into a plurality of sub-problems, solving the sub-problems using an alternating direction multiplication method, obtaining values ​​corresponding to brightness compensation, and further obtaining an underwater image after brightness compensation;

[0011] S4: Obtain the medium transmittance and underwater ambient light of the three channels R, G, and B based on the DCP algorithm;

[0012] S5: Processing the underwater image after brightness compensation using the medium transmittance and the underwater ambient light to obtain a restored image.

[0013] Furthermore, an attenuation model and a brightness correction matrix of the original underwater image from the water surface to the scene are constructed, including:

[0014] S21, constructing an attenuation model of the original underwater image from the water surface to the scene, as shown in formula (1),

[0015] I λ (x) = R λ (x)·T λ (x) (1)

[0016] Among them, R λ (x) is the underwater image without brightness attenuation, T λ (x) is the residual energy ratio of ambient light, I λ (x) is the attenuated image;

[0017] S22. Calculate the components of the brightness channel of the original underwater image in the HSV space, as shown in formulas (2) and (3):

[0018]

[0019] v=max(R′,G′,B′) (3)

[0020] Where R r ,G g ,B b are the matrices of the three channels of the original underwater image in the RGB color space, and v is the component of the brightness channel of the original underwater image in the HSV space;

[0021] S23, constructing a brightness correction matrix according to the components of the brightness channel of the original underwater image in the HSV space.

[0022] Furthermore, a brightness correction matrix is ​​constructed according to the components of the brightness channel of the original underwater image in the HSV space, including:

[0023] S231: Calculate the first component of the brightness correction matrix, as shown in formulas (4) and (5),

[0024] v1=v-lg(1+v) (4)

[0025]

[0026] Where cor1 is the first component of the calculated brightness correction matrix, and v1 represents the difference between v and its logarithmic change;

[0027] S232: Calculate the second component of the brightness correction matrix, as shown in formulas (6) and (7),

[0028]

[0029] Wherein, m and n are the length and width of the original underwater image, σ is a constant, cor2 is the second component of the calculated brightness correction matrix; v2 represents the logarithmic mean of v;

[0030] S233: assign weights to the first component and the second component to obtain a brightness correction matrix, as shown in formula (8):

[0031] c=n×cor1+(1-n)×cor2 (8)

[0032] Among them, n represents the weight.

[0033] Further, S3 defines a brightness compensation function according to the attenuation model and the brightness correction matrix, including:

[0034] S31, defining a brightness compensation function based on the attenuation model, the brightness correction matrix and the augmented Lagrangian function, as shown in formula (9),

[0035]

[0036] Among them, α, β, η are regularization parameters, and the first term is the fidelity term, the second term To minimize the gradient of the decay component, the third term is the brightness compensation term, the fourth term Refers to edge sparsity; R is another way of expressing the underwater image without brightness attenuation, T is another way of expressing the residual energy ratio of ambient light, I is another way of expressing the image after attenuation, and k represents a constant parameter;

[0037] S32, introduce an auxiliary variable u and an error b, and redefine the brightness compensation function, as shown in formula (10),

[0038]

[0039] Further, S3 decomposes the brightness compensation function into multiple sub-problems, and uses the alternating direction multiplier method to solve the sub-problems to obtain values ​​corresponding to the brightness compensation, including:

[0040] S33, decomposing the brightness compensation function into sub-problems about R and T;

[0041] S34. Solve the subproblems using the alternating direction multiplier method. The j-th iteration solutions of subproblems R and T are as shown in formulas (11) and (12).

[0042]

[0043] in Ψ represents the intermediate variable, is a difference operator, applied to the horizontal and vertical directions of the image; F corresponds to the fast Fourier transform, F * Corresponding to the conjugate transpose, F -1 Corresponding to the inverse operation of FFT, ε0 is a constant; R j represents the solution of the jth iteration of the underwater image with unattenuated brightness, T j Represents the remaining energy of the ambient light compared to the solution of the jth iteration.

[0044] Furthermore, S3 obtains the underwater image after brightness compensation, including:

[0045] S35, obtaining the image after brightness compensation, as shown in formula (13),

[0046] R λ (x) = J λ (x)t λ (x)+A λ (1-t λ (x)) (13)

[0047] By shifting the terms in formula (13), we can get formula (14):

[0048]

[0049] Where R λ (x) represents the image without brightness attenuation, that is, the image after brightness compensation, and its value is R j , J λ (x) is the high-quality image to be obtained, t λ (x) is the medium transmittance, A λ It is the underwater ambient light.

[0050] Furthermore, S4 obtains the medium transmittance and underwater ambient light of the three channels R, G, and B based on the DCP algorithm. The medium transmittance and underwater ambient light are used to restore the underwater image to obtain a restored image, including:

[0051] S41. Based on the dark channel prior theory, the minimum value of the dark channel image coordinates is obtained, as shown in formula (15):

[0052]

[0053] Among them, Ω(x,y) represents the self-defined area centered on the pixel point (x,y) in the original underwater image, (u,v) represents any pixel point in the area, and J λ (u,v) represents a high-quality image at any pixel;

[0054] Based on the DCP algorithm, two minimum filters are used to estimate the local transmittance, as shown in formulas (16) and (17),

[0055]

[0056] Among them, I λ (u,v) represents the attenuated image of any pixel;

[0057] S42, substituting formula (15) into formula (17) to obtain the transmittance estimation value based on the DCP algorithm, as shown in formula (18),

[0058]

[0059] S43, respectively estimate the medium transmittance of the r, g, and b channels based on the transmittance estimation formula, as shown in formula (19),

[0060]

[0061] Among them, A λ(u, v) represents the pixel value of the underwater ambient light at any pixel point;

[0062] S54, using quadtree algorithm to estimate underwater ambient light A λ The pixel value of .

[0063] Further, S5 uses the medium transmittance and the underwater ambient light to process the underwater image after brightness compensation to obtain a restored image, including:

[0064] The medium transmittance t λ (x,y) and A λ Substitute it into formula (14) to obtain the restored image.

[0065] Beneficial effects: The present invention provides an underwater image restoration method based on adaptive brightness improvement and local image descattering. By constructing an attenuation model and a brightness correction matrix, the brightness degradation problem of underwater images is paid attention to, so that underwater images are closer to images under real ambient light. At the same time, according to a more objective brightness compensation function of the attenuation model and the brightness correction matrix, the global contrast, brightness and details of the image are fully restored. Different medium transmittances are used to reconstruct the R, G, and B channels of the image instead of using a global medium transmittance, so that the reconstruction result is of higher quality and the descattering is more accurate. A better balance is achieved in processing the blurred contrast and low brightness of different underwater images. The method does not require any fine-tuning of parameters and can well restore underwater images. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0067] Figure 1 A method flow chart of an underwater image restoration method based on adaptive brightness improvement and local image descattering according to the present invention;

[0068] Figure 2 The histogram distribution diagram of the underwater sample image and its R, G, and B channels of the present invention;

[0069] Figure 3 Take model pictures for underwater images;

[0070] Figure 4 An original underwater image and a restored image result diagram according to an embodiment of the present invention;

[0071] Figure 5This is a qualitative test result diagram of an embodiment of the present invention on the UIEB data set;

[0072] Figure 6 This is a quantitative test result diagram of an embodiment of the present invention on the UIEB data set;

[0073] Figure 7 This is a generalization result diagram of remote sensing images and nighttime images according to an embodiment of the present invention. DETAILED DESCRIPTION

[0074] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0075] This embodiment provides an underwater image restoration method based on adaptive brightness improvement and local image descattering. Figure 1 As shown, including:

[0076] S1: Acquire an original underwater image, where the original underwater image is an RGB image;

[0077] S2: constructing an attenuation model and a brightness correction matrix of the original underwater image from the water surface to the scene;

[0078] S3: defining a brightness compensation function according to the attenuation model and the brightness correction matrix, decomposing the brightness compensation function into a plurality of sub-problems, solving the sub-problems using an alternating direction multiplication method, obtaining values ​​corresponding to brightness compensation, and further obtaining an underwater image after brightness compensation;

[0079] S4: Obtain the medium transmittance and underwater ambient light of the three channels R, G, and B based on the DCP algorithm;

[0080] S5: Processing the underwater image after brightness compensation using the medium transmittance and the underwater ambient light to obtain a restored image.

[0081] Specifically, firstly, the original underwater image is obtained through an open source database. The image is an RGB image, which provides material for the subsequent conversion into an HSV format image; secondly, an attenuation model and a brightness correction matrix of the original underwater image from the water surface to the scene are constructed. The attenuation model from the water surface to the scene is constructed, which can remove the energy attenuation caused by the absorption and scattering of suspended particles in the underwater medium, and simulate the image under natural light to construct a brightness matrix, which can compensate for the extinction from the water surface to the scene, improve the problem of low image brightness and color decay, and make the underwater image clearer; the brightness compensation function is defined according to the attenuation model and the brightness correction matrix. , and decompose the brightness compensation function into multiple sub-problems, use the alternating direction multiplication method to solve the sub-problems, obtain the value corresponding to the brightness compensation, and then obtain the underwater image after brightness compensation, which fully restores the global contrast, brightness and details of the image; based on the DCP algorithm, the medium transmittance and underwater ambient light of the three channels R, G, and B are obtained, and different medium transmittances are used to reconstruct the R, G, and B channels of the image instead of using a global medium transmittance, so that the reconstruction result is higher in quality and the descattering is more accurate; finally, the underwater image after brightness compensation is processed using the medium transmittance and the underwater ambient light to obtain a restored image.

[0082] In a specific embodiment, the scheme for obtaining an original underwater image, where the original underwater image is an RGB image, is:

[0083] In this embodiment, the original underwater image is obtained through the open source database UIEB. UIEB is the first large-scale dataset containing real-world underwater images and corresponding reference images for comprehensive evaluation of underwater image enhancement. Figure 2 As shown, the first picture is the original underwater image, the second picture is the image of the R channel, the third picture is the image of the G channel, the fourth picture is the image of the B channel, and the fifth picture is the histogram distribution of the RGB three-channel image.

[0084] In a specific embodiment, the scheme for constructing the attenuation model and brightness correction matrix of the original underwater image from the water surface to the scene is:

[0085] S21, constructing an attenuation model of the original underwater image from the water surface to the scene, as shown in formula (20),

[0086] I λ (x)= R λ(x)·T λ (x) (20)

[0087] Among them, R λ (x) is the underwater image without brightness attenuation, T λ (x) is the residual energy ratio of ambient light, I λ (x) is the attenuated image; Figure 3 This is an underwater shooting model diagram. The left side of the figure is the camera, the right side is the scene, and the particles in the figure are suspended particles in the water. In the figure, L1 represents direct scattering, L2 represents forward scattering, and L3 represents backscattering. It can be seen from the figure that the camera will be affected by the suspended particles in the water when shooting underwater, causing the light that the camera can shoot to have forward scattering and back scattering, which in turn causes the details of the captured image to be blurred and the contrast to be low;

[0088] S22, calculate the components of the brightness channel of the original underwater image in the HSV space, as shown in formulas (21) and (22),

[0089]

[0090] v=max(R′,G′,B′) (22)

[0091] Among them, R r ,G g ,B b are the matrices of the three channels of the original underwater image in the RGB color space, and v is the component of the brightness channel of the original underwater image in the HSV space;

[0092] S23, constructing a brightness correction matrix according to the components of the brightness channel of the original underwater image in the HSV space:

[0093] S231: Calculate the first component of the brightness correction matrix, as shown in formulas (23) and (24),

[0094] v1=v-lg(1+v) (23)

[0095]

[0096] Where cor1 is the first component of the calculated brightness correction matrix, and v1 represents the difference between v and its logarithmic change;

[0097] S232: Calculate the second component of the brightness correction matrix, as shown in formulas (25) and (26),

[0098]

[0099] Wherein, m and n are the length and width of the original underwater image, σ is a constant, which is a very small value. In order to prevent the occurrence of black spots with brightness of 0 in the image, the value is 0.001, cor2 is the second component of the calculated brightness correction matrix; the value of v is a matrix, and in formula (25), the summation formula is solved to find the sum of the pixel logarithms of the matrix v;

[0100] S233: assign weights to the first component and the second component to obtain a brightness correction matrix, as shown in formula (27):

[0101] c=n×cor1+(1-n)×cor2 (27)

[0102] The weights in this embodiment are determined by experimental tests. The value of n is 0.3, that is, the brightness correction matrix c=0.3×cor1+0.7×cor2. Under this value, the brightness matrix has the best effect of improving the image.

[0103] In this scheme, an attenuation model from the water surface to the scene is constructed, which can remove the energy attenuation caused by the absorption and scattering of suspended particles in the underwater medium. The brightness matrix is ​​constructed by simulating the image under natural light, which can compensate for the extinction along the water surface to the scene, improve the problems of low image brightness and color decay, and make underwater images clearer.

[0104] In a specific embodiment, a brightness compensation function is defined according to the attenuation model and the brightness correction matrix, and the brightness compensation function is decomposed into multiple sub-problems, and the sub-problems are solved using the alternating direction multiplication method to obtain the value corresponding to the brightness compensation, and then the solution for obtaining the underwater image after brightness compensation is:

[0105] S31, defining a brightness compensation function based on the attenuation model, the brightness correction matrix and the augmented Lagrangian function, as shown in formula (28),

[0106]

[0107] Among them, αα, β, η are regularization parameters, and the first term is the fidelity term, the second term To minimize the gradient of the decay component, the third term is the brightness compensation term, the fourth term Refers to edge sparsity; R is another way of expressing the underwater image without brightness attenuation, T is another way of expressing the residual energy ratio of ambient light, I is another way of expressing the image after attenuation, and k represents a constant parameter;

[0108] S32, introduce an auxiliary variable u and an error b, and redefine the brightness compensation function, as shown in formula (29),

[0109]

[0110] S33, decomposing the brightness compensation function into sub-problems about R and T, as shown in formulas (30) and (31),

[0111]

[0112] S34. Solve the subproblems using the alternating direction multiplier method. The j-th iteration solutions of subproblems R and T are as shown in formulas (32) and (33).

[0113]

[0114] in represents the intermediate variable, is a difference operator, applied to the horizontal and vertical directions of the image; F corresponds to the fast Fourier transform, F * Corresponding to the conjugate transpose, F -1 Corresponding to the inverse operation of FFT, ε0 is a constant, which is 0.001 in this embodiment, and the value is extremely small to avoid the situation where the denominator is 0; R j represents the solution of the jth iteration of the underwater image with unattenuated brightness, T j The remaining energy of the ambient light is represented by the solution of the jth iteration. The derivative operator is diagonalized after FFT to prevent matrix inversion. All calculations are performed in element-wise order;

[0115] S35, obtaining the image after brightness compensation, as shown in formula (34),

[0116] R λ (x) = J λ (x)t λ (x)+A λ (1-t λ (x)) (34)

[0117] By rearranging formula (34), we get formula (35):

[0118]

[0119] Where R λ (x) represents the image without brightness attenuation, that is, the image after brightness compensation, and its value is R j , J λ (x) is the high-quality image to be obtained, t λ (x) is the medium transmittance, A λ It is the underwater ambient light.

[0120] In this scheme, a brightness compensation function is defined and decomposed into several sub-problems for solution to obtain the compensated image. The brightness degradation problem of underwater images is paid attention to, and the real-world prior is properly integrated. The best preserved color channel prior values ​​are used to construct a more objective brightness compensation function, which fully restores the global contrast, brightness and details of the image.

[0121] The alternating direction multiplier method is an important method for solving separable convex optimization problems. Due to its fast processing speed and good convergence performance, the ADMM algorithm is widely used in statistical learning, machine learning and other fields. This solution uses this method to solve sub-problems. By decomposing the original problem into a series of smaller sub-problems, the solution process is more efficient and easier to manage.

[0122] In a specific embodiment, the solution for obtaining the medium transmittance of the three channels R, G, and B and the underwater ambient light based on the DCP algorithm is:

[0123] S41. Based on the dark channel prior theory, the minimum value of the dark channel image coordinates is obtained, as shown in formula (36):

[0124]

[0125] Among them, Ω(x,y) represents the self-defined area centered on the pixel point (x,y) in the original underwater image, (u,v) represents any pixel point in the area, and J λ (u,v) represents a high-quality image at any pixel;

[0126] Based on the DCP algorithm, two minimum filters are used to estimate the local transmittance, as shown in equations (37) and (38),

[0127]

[0128] Among them, I λ (u,v) represents the attenuated image of any pixel;

[0129] S42, substituting formula (36) into formula (38) to obtain the transmittance estimation value based on the DCP algorithm, as shown in formula (39),

[0130]

[0131] S43, respectively estimate the medium transmittance of the r, g, and b channels based on the transmittance estimation formula, as shown in formula (40),

[0132]

[0133] Among them, A λ (u, v) represents the pixel value of the underwater ambient light at any pixel point;

[0134] S54, using quadtree algorithm to estimate underwater ambient light A λ Pixel value:

[0135] The quadtree hierarchical search method is used to divide the input image into four regions on average, find the region with the smallest variance and the region with the largest color difference, and calculate the average value of the two regions as the background light candidates A1 and A2. At the same time, in order to reflect the depth of field effect of the underwater environment, the scene depth map is obtained by the disparity of the bright and dark channels, and the average value of the R, G, and B channels of the image where the first 0.1% of the pixels in the depth map are located is selected as the background light candidate A3. Finally, an adaptive weight method is used to estimate the underwater ambient light.

[0136] In this scheme, different medium transmittances are used to reconstruct the R, G, and B channels of the image instead of using a global medium transmittance. The reconstruction result is of higher quality and the descattering is more accurate. Independent calculation is used to estimate the medium transmittance of the R, G, and B channels, which can solve the problem that the same transmittance cannot describe the haze level of different channels, seriously affecting the image enhancement quality, so that the enhanced result has residual haze or color cast. At the same time, it solves the problem of low image contrast caused by backscattering and displays more scene details.

[0137] In a specific embodiment, the solution of processing the underwater image after brightness compensation using the medium transmittance and the underwater ambient light to obtain the restored image is:

[0138] At the known parameter t λ (x) and A λ On the basis of, substitute into formula (35) to get the restored image.

[0139] Embodiment 1:

[0140] The original underwater image is processed by this method. Figure 4 The result of restoring the first line is as follows Figure 4 The second row is shown.

[0141] Embodiment 2:

[0142] This method and other existing methods are used to analyze underwater images. Figure 5 (a) Perform restoration, and the result is as follows Figure 5 As shown:

[0143] (a) is the original underwater image, (b) is the output result of DCP, (c) is the output result of DAACC, (d) is the output result of GDCP, (e) is the output result of MIP, (f) is the output result of Red-channel, (g) is the output result of NUDCP, (h) is the output result of DATVR, (i) is the output result of UDCP, (j) is the output result of CBM, (k) is the output result of ULAP, (l) is the output result of Waternet, (m) is the output result of FuniGAN, and (n) is the output result of this embodiment.

[0144] Embodiment 3:

[0145] This method and other existing methods are used to analyze underwater images. Figure 6 (a) Restore and compare, the result is as follows Figure 6 As shown:

[0146] (a) is the original underwater image, (b) is the output result of DCP, (c) is the output result of DAACC, (d) is the output result of GDCP, (e) is the output result of MIP, (f) is the output result of Red-channel, (g) is the output result of NUDCP, (h) is the output result of DATVR, (i) is the output result of UDCP, (j) is the output result of CBM, (k) is the output result of ULAP, (l) is the output result of Waternet, (m) is the output result of FuniGAN, and (n) is the output result of this embodiment.

[0147] Embodiment 4:

[0148] This method is used to compare underwater image restoration data with other existing methods. The average quantitative results obtained by executing different methods on the UIEB dataset are shown in Table 1.

[0149] Table 1

[0150]

[0151]

[0152] The data in Table 1 show that the Blur, DFAD and UCIQE evaluation indicators of the method of the present invention are better than those of other methods. The method of the present invention has a significant improvement effect on the brightness, saturation and contrast of the image. After being processed by the method of the present invention, the quality of the underwater image is significantly improved.

[0153] This method has good generalization ability for processing remote sensing and nighttime images, can eliminate haze, and improve the brightness and contrast of remote sensing and nighttime images, as shown in Example 5:

[0154] This method is used to restore remote sensing images and nighttime images, and the results are as follows: Figure 7 As shown:

[0155] From left to right are two remote sensing images and two night images. From the image quality, it can be seen that the use of this method can achieve the effect of defogging and improving brightness and contrast.

[0156] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An underwater image restoration method based on adaptive brightness improvement and local image descattering, It is characterized in that include: S1: Acquire an original underwater image, where the original underwater image is an RGB image; S2: constructing an attenuation model and a brightness correction matrix of the original underwater image from the water surface to the scene; S3: defining a brightness compensation function according to the attenuation model and the brightness correction matrix, decomposing the brightness compensation function into a plurality of sub-problems, solving the sub-problems using an alternating direction multiplication method, obtaining values ​​corresponding to brightness compensation, and further obtaining an underwater image after brightness compensation; S4: Obtain the medium transmittance and underwater ambient light of the three channels R, G, and B based on the DCP algorithm; S5: Processing the underwater image after brightness compensation using the medium transmittance and the underwater ambient light to obtain a restored image.

2. The underwater image restoration method based on adaptive brightness improvement and local image descattering according to claim 1, It is characterized in that S2 constructs an attenuation model and a brightness correction matrix of the original underwater image from the water surface to the scene, including: S21, constructing an attenuation model of the original underwater image from the water surface to the scene, as shown in formula (1), I λ (x)=R λ (x)·T λ (x) (1) Among them, R λ (x) is the underwater image without brightness attenuation, T λ (x) is the residual energy ratio of ambient light, I λ (x) is the attenuated image; S22, calculate the components of the brightness channel of the original underwater image in the HSV space, as shown in formulas (2) and (3), v=max(R′,G′,B′) (3) Among them, R r ,G g ,B b are the matrices of the three channels of the original underwater image in the RGB color space, and v is the component of the brightness channel of the original underwater image in the HSV space; S23, constructing a brightness correction matrix according to the components of the brightness channel of the original underwater image in the HSV space.

3. The underwater image restoration method based on adaptive brightness improvement and local image descattering according to claim 2, It is characterized in that A brightness correction matrix is ​​constructed according to the components of the brightness channel of the original underwater image in the HSV space, including: S231: Calculate the first component of the brightness correction matrix, as shown in formulas (4) and (5), v 1 =v-lg(1+v) (4) Among them, cor 1 is the first component of the calculated brightness correction matrix, v 1 Represents the difference between v and its logarithmic change; S232: Calculate the second component of the brightness correction matrix, as shown in formulas (6) and (7), Where m and n are the length and width of the original underwater image, σ is a constant, cor 2 is the second component of the calculated brightness correction matrix; v 2 represents the logarithmic mean of v; S233: assign weights to the first component and the second component to obtain a brightness correction matrix, as shown in formula (8): v=n×cor 1 +(1-n)×cor 2 (8) Among them, n represents the weight.

4. The underwater image restoration method based on adaptive brightness improvement and local image descattering according to claim 1, It is characterized in that S3 defines a brightness compensation function according to the attenuation model and the brightness correction matrix, including: S31, defining a brightness compensation function based on the attenuation model, the brightness correction matrix and the augmented Lagrangian function, as shown in formula (9), Among them, α, β, η are regularization parameters, and the first term is the fidelity term, the second term To minimize the gradient of the decay component, the third term is the brightness compensation term, the fourth term Refers to edge sparsity; R is another way of expressing the underwater image without brightness attenuation, T is another way of expressing the residual energy ratio of ambient light, I is another way of expressing the image after attenuation, and k represents a constant parameter; S32, introduce an auxiliary variable u and an error b, and redefine the brightness compensation function, as shown in formula (10), 5. The underwater image restoration method based on adaptive brightness improvement and local image descattering according to claim 1, It is characterized in that S3 decomposes the brightness compensation function into multiple sub-problems, and uses the alternating direction multiplier method to solve the sub-problems to obtain values ​​corresponding to the brightness compensation, including: S33, decomposing the brightness compensation function into sub-problems about R and T; S34. Solve the subproblems using the alternating direction multiplier method. The j-th iteration solutions of subproblems R and T are as shown in formulas (11) and (12). in Ψ represents the intermediate variable, is a difference operator, applied to the horizontal and vertical directions of the image; F corresponds to the fast Fourier transform, F * Corresponding to the conjugate transpose, F -1 Corresponding to the inverse operation of FFT, ε 0 is a constant; R j represents the solution of the jth iteration of the underwater image with unattenuated brightness, T j Represents the remaining energy of the ambient light compared to the solution of the jth iteration.

6. According to the underwater image restoration method based on adaptive brightness improvement and local image descattering according to claim 5, S3 obtains the underwater image after brightness compensation, include: S35, obtaining the image after brightness compensation, as shown in formula (13), R λ (x)=J λ (x)t λ (x)+A λ (1-tλ(x)) (13) By shifting the terms in formula (13), we can get formula (14): Where R λ (x) represents the image without brightness attenuation, that is, the image after brightness compensation, and its value is R j , J λ (x) is the high-quality image to be obtained, t λ (x) is the medium transmittance, A λ It is the underwater ambient light.

7. The underwater image restoration method based on adaptive brightness improvement and local image descattering according to claim 1, It is characterized in that S4 obtains the medium transmittance and underwater ambient light of the three channels R, G, and B based on the DCP algorithm, including: S41. Based on the dark channel prior theory, the minimum value of the dark channel image coordinates is obtained, as shown in formula (15): Among them, Ω(x,y) represents the self-defined area centered on the pixel point (x,y) in the original underwater image, (u,v) represents any pixel point in the area, and J λ (u,v) represents a high-quality image at any pixel; Based on the DCP algorithm, two minimum filters are used to estimate the local transmittance, as shown in formulas (16) and (17), Among them, I λ (u,v) represents the attenuated image of any pixel; S42, substituting formula (15) into formula (17) to obtain the transmittance estimation value based on the DCP algorithm, as shown in formula (18), S43, respectively estimate the medium transmittance of the r, g, and b channels based on the transmittance estimation formula, as shown in formula (19), Among them, A λ (u, v) represents the pixel value of the underwater ambient light at any pixel point; S54, using quadtree algorithm to estimate underwater ambient light A λ The pixel value of .

8. The underwater image restoration method based on adaptive brightness improvement and local image descattering according to claim 7, It is characterized in that S5 uses the medium transmittance and the underwater ambient light to process the underwater image after brightness compensation to obtain a restored image, including: The medium transmittance t λ (x,y) and A λ Substitute it into formula (14) to obtain the restored image.

Images